def load_data(test_size=0.2, from_npy=True):
if from_npy:
X = np.load('/export/home/lparcala/dataFace/Xsaved.npy')
y = np.load('/export/home/lparcala/dataFace/ysaved.npy')
else:
y = readLabels('wanted_attributes.txt', 'list_attr_celeba.txt')
X = np.array(ImageSequence('../Face/img_align_celeba.zip'))
X = X.astype('float32')
# convert labels to boolean
y[y < 0] = 0
y[y > 0] = 1
h5f = h5py.File('data.h5', 'w')
h5f.create_dataset('images', data=X)
h5f.create_dataset('labels', data=y)
# np.save("/export/home/lparcala/dataFace/Xsaved.npy", X)
np.save("/export/home/lparcala/dataFace/ysaved.npy", y)
print("Done reading data from disk.")
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=test_size,
random_state=42)
return X_train, y_train, X_test, y_test
def _load_data(self, filepath: Union[str, Path]):
if type(filepath) == str:
filepath = Path(filepath)
if not filepath.is_dir():
raise ValueError(
f'The path you provided does not exist: {filepath}.')
paths = Path(filepath)
image_paths = list(paths.glob("*.tif"))
try:
meta_data_path = list(paths.glob("*.txt"))[0]
except IndexError:
raise ValueError(
f'Could not find a metadata file in {filepath}. Expected to find a .txt file there.'
)
if not image_paths:
raise ValueError(
f"Could not find any tif image files in {filepath}")
with meta_data_path.open("r", encoding="latin-1") as fid:
text_file = fid.readlines()
for line in text_file:
if VOXEL_HEIGHT in line.split():
self.meta_data[VOXEL_HEIGHT] = float(line.split()[-1][:-1])
if VOXEL_HEIGHT not in self.meta_data:
raise ValueError(
f'Unable to find voxel height in {meta_data_path}')
images = ImageSequence(list(image_paths), as_grey=True)
grey_images = self._get_grey_sequence(images)
self._threshold_and_crop_images(grey_images)
def prepare(self):
directory = os.path.join(path, 'video', 'image_sequence')
self.v = tp.invert_image(ImageSequence(os.path.join(directory, '*.png')))
# mass depends on pixel dtype, which differs per reader
minmass = self.v[0].max() * 2
self.PARAMS = {'diameter': 11, 'minmass': minmass}
self.expected = tp.batch(self.v[[0, 1]], engine='python', meta=False,
**self.PARAMS)
def correctImagesCFG(home, imgFolder, imgExt, selectBox, autoBox, crop, stds):
targetDir = os.path.join(home, imgFolder)
outDir = os.path.join(targetDir, "corrected")
fullImg = 1
if not os.path.exists(outDir):
os.makedirs(outDir)
filePattern = os.path.join(targetDir, "*." + imgExt)
first = 1
for filename in glob.glob(filePattern):
images = ImageSequence(filename)
imgcv = images[0]
imName = os.path.basename(filename)
prePost = imName.split(".")
noEnd = prePost[0]
if selectBox:
if first == 1:
fig, ax = plt.subplots()
imgplot = ax.imshow(imgcv)
getter1 = getter(ax, fig)
plt.show()
x = getter1.x
y = getter1.y
x.sort()
y.sort()
for i in range(0, len(x)):
x[i] = int(x[i])
for i in range(0, len(y)):
y[i] = int(y[i])
first = 0
elif autoBox:
x = crop[0]
y = crop[1]
else:
dims = imgcv.shape
x = [0, dims[0]]
y = [0, dims[1]]
croppedImg = imgcv[y[0]:y[1], x[0]:x[1]]
outImg = standardizeCropped(imgcv, croppedImg, stds)
outImg = outImg * 256
outImg = outImg.astype(np.uint8)
im = Image.fromarray(outImg)
saveName = os.path.join(outDir, imName)
im.save(saveName)
def run():
# download to this folder
current_folder = os.sep.join(__file__.split(os.sep)[:-1])
dpc_demo_data_path = os.path.join(current_folder, 'SOFC')
if not os.path.exists(dpc_demo_data_path):
sofc_file = os.path.join(current_folder, 'SOFC.zip')
print('The required test data directory was not found.'
'\nDownloading the test data to %s' % dpc_demo_data_path)
# todo make this not print every fraction of a second
call(('wget https://www.dropbox.com/s/963c4ymfmbjg5dm/SOFC.zip -P %s' %
current_folder),
shell=True)
# unzip it into this directory
unzip(sofc_file)
# 1. Set parameters
start_point = [1, 0]
first_image = 1
pixel_size = (55, 55)
focus_to_det = 1.46e6
scan_xstep = 0.1
scan_ystep = 0.1
scan_rows = 121
scan_cols = 121
energy = 19.5
roi = None
padding = 0
weighting = 1.
bad_pixels = None
solver = 'Nelder-Mead'
images = ImageSequence(dpc_demo_data_path + "/*.tif")
img_size = images[0].shape
ref_image = np.ones(img_size)
scale = True
negate = True
print('running dpc')
# 2. Use dpc.dpc_runner
phase, amplitude = dpc.dpc_runner(ref_image, images, start_point,
pixel_size, focus_to_det, scan_rows,
scan_cols, scan_xstep, scan_ystep,
energy, padding, weighting, solver, roi,
bad_pixels, negate, scale)
# 3. Save intermediate and final results
scipy.misc.imsave(os.path.join(current_folder, 'phase.jpg'), phase)
np.savetxt(os.path.join(current_folder, 'phase.txt'), phase)
scipy.misc.imsave(os.path.join(current_folder, 'amplitude.jpg'), amplitude)
np.savetxt(os.path.join(current_folder, 'amplitude.txt'), amplitude)
def build_image_cycle(path):
"""Build image cycles, essentially generators with endless images
Parameters
----------
path: str
Path to the files to be used as the base for the cycle, this can
include some globing
Returns
-------
Cycler:
The iterable like object to cycle through the images
"""
imgs = ImageSequence(os.path.join(path, '*.tif*'))
return cycler(pe1_image=[i for i in imgs])
def open(self, initializer: Path = None):
if initializer is not None:
self.initializer = initializer
if not self.initializer.is_dir():
raise ValueError(f'{self.initializer} must be a directory')
self.imseq = ImageSequence(str(self.initializer))
setattr(self, 'close', getattr(self.imseq, 'close'))
self.nfrms = len(self.imseq)
self.ysize = self.imseq.frame_shape[0]
self.xsize = self.imseq.frame_shape[1]
self.dtype = self.imseq._dtype
if len(self.imseq.frame_shape) > 2:
self.nchannels = self.imseq.frame_shape[2]
def run():
# download to this folder
current_folder = os.sep.join(__file__.split(os.sep)[:-1])
dpc_demo_data_path = os.path.join(current_folder, 'SOFC')
if not os.path.exists(dpc_demo_data_path):
zip_file_url = 'https://www.dropbox.com/s/963c4ymfmbjg5dm/SOFC.zip?dl=1'
import download
download.run()
# download_zip(zip_file_url, current_folder)
# 1. Set parameters
start_point = [1, 0]
first_image = 1
pixel_size = (55, 55)
focus_to_det = 1.46e6
scan_xstep = 0.1
scan_ystep = 0.1
scan_rows = 121
scan_cols = 121
energy = 19.5
roi = None
padding = 0
weighting = 1.
bad_pixels = None
solver = 'Nelder-Mead'
images = ImageSequence(dpc_demo_data_path + "/*.tif")
img_size = images[0].shape
ref_image = np.ones(img_size)
scale = True
negate = True
print('running dpc')
# 2. Use dpc.dpc_runner
phase, amplitude = dpc.dpc_runner(ref_image, images, start_point,
pixel_size, focus_to_det, scan_rows,
scan_cols, scan_xstep, scan_ystep,
energy, padding, weighting, solver, roi,
bad_pixels, negate, scale)
# 3. Save intermediate and final results
print('saving dpc output to disk in --> %s' % current_folder)
scipy.misc.imsave(os.path.join(current_folder, 'phase.jpg'), phase)
np.savetxt(os.path.join(current_folder, 'phase.txt'), phase)
scipy.misc.imsave(os.path.join(current_folder, 'amplitude.jpg'), amplitude)
np.savetxt(os.path.join(current_folder, 'amplitude.txt'), amplitude)
from skimage.segmentation import mark_boundaries
from skimage.util import img_as_float
from sklearn.preprocessing import Normalizer
from sklearn import metrics
from sklearn.cluster import KMeans, MiniBatchKMeans
from mpl_toolkits.mplot3d import Axes3D
import matplotlib.pyplot as plt
import warnings
warnings.filterwarnings('ignore')
# ## 1-loading Frames And Masks Into Program
# In[3]:
images = ImageSequence('images//birdfall2_000*.png')
masks = ImageSequence('masks//birdfall2_000*.png')
# In[34]:
framecount=29
fig = plt.figure(figsize=(10, 140))
fig.suptitle('20 Object Porposals For 20 Frames')
for k in range(0,framecount):
img=images[k]
ax = fig.add_subplot(framecount, 3, k+1)
ax.imshow(img)
self.fig.canvas.draw()
targetDir = os.path.join(os.getcwd(), 'annotations')
ext = 'tif'
outDir = targetDir + "\\corrected\\"
fullImg = 1
if not os.path.exists(outDir):
os.makedirs(outDir)
filePattern = targetDir + "\\*." + ext
first = 1
for filename in glob.glob(filePattern):
#imgcv = cv2.imread(filename)
images = ImageSequence(filename)
imgcv = images[0]
print(imgcv.dtype)
fig, ax = plt.subplots()
imgplot = ax.imshow(imgcv)
sections = filename.split("\\")
imName = sections[-1]
#im.save(outDir+imName)
#print(result)
prePost = imName.split(".")
noEnd = prePost[0]
if first == 1:
getter1 = getter(ax, fig)
#define model function and pass independant variables x and y as a list
def twoD_Gaussian(x_data_tuple, amplitude, xo, yo, sigma_x, sigma_y,
offset):
(x, y) = x_data_tuple
xo = float(xo)
yo = float(yo)
g = offset + amplitude * np.exp(-(((x - xo) / sigma_x)**2 +
((y - yo) / sigma_y)**2) / 2)
return g.ravel()
try:
os.stat('fit_images')
except:
os.mkdir('fit_images')
im = ImageSequence('./*.tif')
im_height, im_width = im[0].shape
bound_tup = ([15, 5, 5, 0, 0, 0],
[+np.Inf, (im_height - 5), (im_width - 5), 1.6, 1.6, +np.Inf])
df_add = []
min_pix = np.min(im)
max_pix = np.max(im)
for cur_frame in im:
# Open Image
print('Frame: ' + str(cur_frame.frame_no))
data = cur_frame
# Create x and y indices
# In[19]:
get_ipython().magic('ls')
# In[20]:
from pims import ImageSequence
# In[21]:
images = ImageSequence('fc2_save_2019-05-23-094552-*.jpg')
# In[22]:
images
# In[23]:
red_images = images[::2]
# In[24]:
Created on Wed Nov 28 13:51:13 2018
@author: hschriever
"""
import numpy as np
import cv2
from pims import ImageSequence
#Import the data and convert to np array
#input from command line
DIR_NAME = input('what is the path?\n')
NUM_EIGENVECTORS = 3
raw_data = np.array(ImageSequence(DIR_NAME))
size = raw_data.shape
image_size = (size[2], size[3])
data_PCA_format = imagesToRows(raw_data) #depth assumed to be in dim 2
mean, eigen_vectors = cv2.PCACompute(data_PCA_format,
mean=None,
maxComponents=NUM_EIGENVECTORS)
#initializing lists that will contain the eigan images
eigen_images = []
eigen_images_pos = []
eigen_images_neg = []
'''
Creating the eigan images
def remove_zeros(arr):
new_arr = []
for i in range(len(arr)):
if(arr[i] > 0):
new_arr.append(arr[i])
return new_arr
#parent dirrectory of the ome files (ie, one fish)
directory = '/media/lauderdale/mnt/remote_servers/data1/image_data/LightSheetMicroscope/20170803/*.tif'
print('starting analysis of ' + directory)
data = ImageSequence(directory)
print('Loaded in the data')
num_ome = len(data)
video = data[0]
frame = video[100]
img = np.array(frame[0])
print_img(img)
img = np.array(frame[1])
print_img(img)
intensities = np.zeros(num_ome * 512)
print('entering for loop')
for i in range(num_ome):
print('working on file: ' + str(i))
def correctImagesCFG(cfg):
targetDir = os.path.join(cfg['temp']['rootDir'],
cfg['imageCorrection']['imgFolder'])
ext = cfg['imageCorrection']['imgExt']
outDir = os.path.join(targetDir, "corrected")
fullImg = 1
if not os.path.exists(outDir):
os.makedirs(outDir)
filePattern = os.path.join(targetDir, "*." + ext)
first = 1
print(filePattern)
for filename in glob.glob(filePattern):
images = ImageSequence(filename)
imgcv = images[0]
#print(imgcv.dtype)
#sections = filename.split("\\")
#imName = sections[-1]
imName = os.path.basename(filename)
prePost = imName.split(".")
noEnd = prePost[0]
if cfg['imageCorrection']['selectBox']:
if first == 1:
fig, ax = plt.subplots()
imgplot = ax.imshow(imgcv)
getter1 = getter(ax, fig)
plt.show()
x = getter1.x
y = getter1.y
x.sort()
y.sort()
for i in range(0, len(x)):
x[i] = int(x[i])
for i in range(0, len(y)):
y[i] = int(y[i])
first = 0
elif cfg['imageCorrection']['autoBox']:
x = cfg['imageCorrection']['cropX']
y = cfg['imageCorrection']['cropY']
else:
dims = imgcv.shape
x = [0, dims[0]]
y = [0, dims[1]]
croppedImg = imgcv[y[0]:y[1], x[0]:x[1]]
stds = cfg['standard']['stds']
outImg = standardizeCropped(imgcv, croppedImg, stds)
outImg = outImg * 256
outImg = outImg.astype(np.uint8)
im = Image.fromarray(outImg)
saveName = os.path.join(outDir, imName)
im.save(saveName)
from skimage.util import img_as_float
from sklearn.preprocessing import Normalizer
from sklearn import metrics
from sklearn.cluster import KMeans, MiniBatchKMeans
from mpl_toolkits.mplot3d import Axes3D
import matplotlib.pyplot as plt
import warnings
warnings.filterwarnings('ignore')
# ## 1-loading Frames And Masks Into Program
# In[3]:
DAGSeg = np.load('Masks/birdfall_masksDagSeg.npy')
KeySeg = np.load('Masks/birdfall_masksKeySeg.npy')
SaliencySeg = ImageSequence('Masks/SaliencySeg//birdfall2_00*.bmp')
SeamSeg = np.load('Masks/birdfall_masksSeamSeg.npy')
JOTSeg = np.load('Masks/birdfall_masksFromJOTS.npy')
GroundTruth = np.load('Masks/birdfall_masksGroundtruth.npy')
# In[4]:
framecount = 20
fig = plt.figure(figsize=(9, 40))
plt.suptitle("GroundTruth", size=36)
for k in range(0, framecount):
img = GroundTruth[:, :, k]
ax = fig.add_subplot(framecount, 5, k + 1)
ax.set_yticklabels([])
ax.set_xticklabels([])
ax.imshow(img, cmap=plt.get_cmap('gray'))